Browse Books

Reviewer: Vladimir Botchev

The market is almost flooded with introductory digital signal processing (DSP) texts and software, ranging from junk to real gems. More than 20 years ago, there were about half a dozen DSP texts, none of them introductory; those books are still perfect in every respect. With DSP everywhere and the crowd of more or less technically oriented users growing, the need for introductory texts has increased. Smith's new book is probably one of the best, although not the most popular. It presents down-to-earth explanations of almost every topic mentioned, although I think the last four, mathematically oriented, chapters should have been placed at appropriate points earlier in the book to reinforce these explanations. Potential readers undoubtedly vary in their interests and background, and I advise more technically inclined readers to begin with those four chapters. Chapter 28, on complex numbers, is a nice introduction to that topic. Although the chapter on Fourier transforms seems like a prerequisite for chapter 29, the two can be read concurrently. Chapter 30, on the Laplace transform, and chapter 31, on the Z transform, familiar to every electrical engineering student, are well designed, with numerous graphics and flowcharts to help readers grasp the concepts. The main body of the text is divided into four parts. The first part, “Foundations,” has four chapters. Chapter 1 attempts to motivate the study of DSP with examples drawn from many fields. Chapter 2 presents a compact introduction to the necessary probability and statistics. Chapter 3 discusses the tools needed for supplying the DSP software with the proper data, representative of the continuous processes to be analyzed; analog-to-digital and digital-to-analog converters are analyzed in some detail. Chapter 4 gives an overview of crucial DSP issues, including number formats, precision, and programming considerations. The choice of BASIC to convey algorithm structure and implementation is a weakness, since modern DSP is increasingly focused on C. However, if readers are willing to deal with BASIC conventions, the numerous programs in the book could be useful as detailed diagrams of the algorithms. The next part, “Fundamentals,” has nine chapters. Chapter 5 introduces linear systems and their basic properties, and tells why sine waves are so common in the study of such systems. Chapters 6 and 7 deal with convolution, and contain one of the most intuitive graphical approaches I have seen in an introductory DSP book. In chapter 8, correlation is explained using the same approach. The next five chapters are devoted to the Fourier transform, its properties, and some of its most common applications. Readers should consider turning to chapters 28 and 29 at this point. Some topics, including the fast Fourier transform (FFT), are not given the treatment I prefer, but this material may suffice as an introduction. Here the weaknesses of BASIC are more obvious, since the FFT structure chosen is not as clear as it would be in C. The last chapter in this part, chapter 13, introduces the concepts of continuous signal processing and should be read along with chapter 30. The next part, “Digital Filters,” contains eight chapters. Chapter 14 introduces digital filter terminology, filter classes, and so on. Chapter 15 intuitively develops the notion of the usefulness of digital filters by discussing the simplest type, the moving average filter. Chapter 16 shows the simplest way to design nonrecursive filters. Chapter 17 is very useful, introducing arbitrary frequency response filters, deconvolution, and optimal filters. Chapter 18 presents FFT-based convolution. Chapter 19 introduces the concepts of poles, zeroes, and recursive filters. This chapter should be read along with chapter 31. Chapter 20 continues with recursive filters, introducing classical types such as Butterworth and Chebyshev; gives tables to help with the design; and presents a BASIC program for the design of Chebyshev filters. The last chapter in this part, chapter 21, compares filters based on their implementation (analog versus digital), types (recursive versus nonrecursive), and so on. The last main part, “Applications,” has six chapters. Chapter 22 introduces digital audio representation and processing. The next three chapters deal with image processing. They are extremely well written, and could by themselves motivate the study of DSP. Chapter 23 discusses image formation and display, explaining brightness and contrast in the more general framework of grayscale transforms. Warping and bilinear interpolation, which are usually considered advanced issues, are also discussed here. Chapter 24 introduces linear image processing, covering edge detection, Fourier image analysis, and convolution processing. Chapter 25, “Special Imaging Techniques,” combines a presentation of the simplest morphological operations with the most readable available introduction to key aspects of computed tomography. Chapter 26 is about neural networks. While suffering a bit from the use of BASIC, this chapter contains something that is rare in DSP books: the design of an arbitrary frequency response digital filter, a perfect example of neural network function approximation. Chapter 27 delves into data compression. While lossy compression is only touched upon, lossless compression such as run-length and Huffman is given more consideration, and the presentation of Lempel-Ziv-Welch (LZW) compression includes implementation details. Chapters 28 through 31, discussed above, make up the final part, “Complex Techniques.” This introductory DSP text is one of the best. I highly recommend it for readers who need a firm understanding of what DSP is and how it can fit their needs.